Cell atlas of the regenerating human liver after portal vein embolization

Abstract

The liver has the remarkable capacity to regenerate. In the clinic, regeneration is induced by portal vein embolization, which redirects portal blood flow, resulting in liver hypertrophy in locations with increased blood supply, and atrophy of embolized segments. Here, we apply single-cell and single-nucleus transcriptomics on healthy, hypertrophied, and atrophied patient-derived liver samples to explore cell states in the regenerating liver. Our data unveils pervasive upregulation of genes associated with developmental processes, cellular adhesion, and inflammation in post-portal vein embolization liver, disrupted portal-central hepatocyte zonation, and altered cell subtype composition of endothelial and immune cells. Interlineage crosstalk analysis reveals mesenchymal cells as an interaction hub between immune and endothelial cells, and highlights the importance of extracellular matrix proteins in liver regeneration. Moreover, we establish tissue-scale iterative indirect immunofluorescence imaging for high-dimensional spatial analysis of perivascular microenvironments, uncovering changes to tissue architecture in regenerating liver lobules. Altogether, our data is a rich resource revealing cellular and histological changes in human liver regeneration.

Fig. 1. Fresh single-cell and frozen single-nucleus RNA-seq reveal major cell populations in adult healthy liver specimens.

a Single-cell and single-nucleus RNA-seq experiments were performed on patient-derived fresh (n = 3) and frozen (n = 3) liver tissues. Prior to single-cell RNA-seq, cells isolated from the fresh tissue were partitioned into parenchymal (hepatocytes) and non-parenchymal (other hepatic cell types) fractions (left). Prior to single-nucleus RNA-seq, nuclei from snap-frozen tissues were isolated using fluorescence activated cell sorting (right). b UMAP plot of transcriptomes from fresh (left) and frozen (right) tissue datasets colored by major cell type (see Supplementary Figs. 1, 2 for each donor’s representation). Gray represents potential doublets. c Scaled expression of marker genes per major cell type across fresh and frozen tissue data (cells and nuclei are represented in rows, genes in columns). d Number of detected genes per cell (top) and nucleus (bottom) across major cell types. Ch cholangiocytes, En endothelial cells, He hepatocytes, Im immune cells, Me mesenchymal cells. Boxplot shows the median (center line), 25th, and 75th percentile (lower and upper boundary), and the whiskers indicate the minimum and maximum values. e Comparison of marker gene detection is shown for each major cell type in fresh and frozen samples (dark gray represents genes shared between both experimental setups; PCC, Pearson correlation coefficient). Source data are provided as a Source Data file for (b–e).

Fig. 2. Transcriptional landscape of human liver cells after portal vein embolization (PVE).

a Schematic shows the portal vein embolization (PVE) procedure and tissue sampling for the scRNA-seq experiments. Portal vein branching toward diseased liver tissue (left lobe) prior to resection is blocked or embolized. Adjacent tissue with redirected blood flow (right lobe) expands over time. ScRNA-seq is performed on two samples derived from regenerating and embolized liver tissues on the day of liver resection. b Computed tomography (CT) scans of pre-PVE (left) and post-PVE (middle) livers are shown along with 3D tissue reconstructions (right), respectively. UMAP plot of merged healthy (n = 3), regenerating (n = 6) and embolized (n = 6) fresh liver samples, colored by condition (c) and major cell type (d). e Major cell type proportion per donor and condition. f Number of differentially expressed genes between healthy and regenerating or embolized samples per major cell type. Gene expression log fold change for each PVE condition compared to healthy (x-axis: embolized; y-axis: regenerating) for cholangiocytes (g), Endothelial cells (h), Hepatocytes (i), and Immune cells (j). k–n, Enriched gene ontology terms for DE genes per major cell type, comparing regenerating or embolized to healthy. Top 6 terms were selected for each condition (two-sided hypergeometric test, dashed line shows Benjamini-Hochberg adjusted p = 0.05). Source data are provided as a Source Data file for (c, d–n).

Fig. 3. Spatial zonation patterns are altered in regenerating and embolized tissue hepatocytes.

a Schematic illustrating the structure of a liver lobule. b 3,3′-Diaminobenzidine (DAB) stainings of central and portal zone-specific protein expression within healthy tissue hepatocytes. c Gene expression signature of zonation within healthy (right), regenerating (middle), and embolized (left) tissue hepatocytes. Cells in UMAP plots are colored based on the cumulative expression of central (blue) or portal (orange) zonation marker genes. d Pseudozonation expression patterns of representative zonation marker genes for each medical condition. e Histograms showing the proportion of hepatocytes in zonation bins across conditions. f Immunofluorescence tissue staining for liver tissue for various hepatocyte zonation markers. P portal vessel, C central vessel. See also Supplementary Figs. 9, 10. g, Top: Representative H/E stainings showing liver lobule segmentation to measure cell density along a central-to-portal axis; Bottom: Mean ± s.e. for cell density along the central-to-portal axis in each condition. Source data are provided as a Source Data file for (d, e).

Fig. 4. Liver endothelial cell heterogeneity and inferred zonation after PVE.

a Schematic illustrates diversity of endothelial cells along the portal-central axis within a liver lobule. UMAP plots representing combined ECs from all medical conditions are colored by condition (b) or annotated subpopulation (c) (EC endothelial cells, LSECs liver sinusoidal endothelial cells, fenestr. fenestrated, MT mitochondrial). d Violin plots show distribution of normalized marker gene expression for endothelial cells subpopulations. e Proportion of each endothelial subpopulation across conditions (healthy, left; regenerating, middle; embolized, right). f Histogram shows comparison of LSECs (periportal, midzonal, and pericentral LSECs) pseudozonation across conditions. LSECs from regenerating and embolized conditions were projected onto a reference healthy zonation trajectory. g Distribution of each LSEC subpopulation from healthy, regenerating and embolized condition is shown along the pseudozonation trajectory. Source data are provided as a Source Data file for (b, c–g).

Fig. 5. Intercellular signaling map reveals fibroblast and immune-cell coordination of response to PVE.

UMAP plots for all immune cells (a), and the myeloid (b) and lymphoid (c) subsets, colored by the respective annotations. d Changes in proportions for select immune cell populations. Gray point shows mean and standard error intervals. * denotes a significant difference compared to healthy (two-tailed binomial test, p < 0.05). H healthy, R regenerating, E embolized. e Scar-associated macrophage (SAMac) signature scores in myeloid cell populations, in each condition. f Number of interactions involving each major cell type, in each condition and in total (left dotplot), along with variation in the number of interactions between healthy and regenerating (middle) or embolized (right). g UMAP plot showing all and condition-specific ligand-receptor interaction networks, summarized by cell type. h Heatmaps showing mean expression for selected ligand-receptor pairs involving ECM proteins across healthy, regenerating, and embolized conditions. Black outline denotes the cell type uniquely expressing the ligand. Gray sidebar distinguishes expression of ligand and receptor. Additional heatmaps are shown in Supplementary Fig. 15. i Regions of Interest from immunohistochemistry slides for two selected ECM proteins. j Total fluorescence intensity of the three proteins assessed in each immunohistochemistry slide, normalized by total DAPI intensity in the tissue area. Source data are provided as a Source Data file for (a–h).

Fig. 6. Multiplexed immunohistochemistry reveals peri-vessel microenvironment alterations in post-PVE patient tissue sections.

a Schematic illustrating the tissue-level 4i protocol. b DAPI staining of healthy liver tissue, illustrating the image resolution across length scales. Multiple staining overlays (expression of CLEC4M, ACTA2, IGFBP7, GLUL, Catalase representing pericentral LSECs, mesenchymal cells, periportal LSECs, hepatocytes, and peroxisomes, respectively) in a large healthy liver tissue section (c) and selected highlighted regions (d), where (i) shows an annotated liver lobule (P portal, C central), and the bottom panels a close-up of a portal (ii) and central (iii) vessel. Multiple staining overlays in large and highlighted sections of embolized (e) and regenerating (f) liver samples. g Portal and central vessel detection results from liver 4i data (Methods, Supplementary Fig. 16). Boxplots (bottom right) show normalized expression of ACTA2 per vessel and condition. Boxplot shows the median (center line), 25th, and 75th percentile (lower and upper boundary), and the whiskers indicate the 1.5 × inter-quartile range, with outliers shown as individual data points. h UMAP plots based on 4i data and derived physical measurements for nuclei segmented in each condition, colored by cell type. i Representative portal and central vessels from each condition, with segmented nuclei colored by cell type. j Variation in cell type proportions shown as normalized proportion of nuclei per vessel area in portal and central vessel microenvironments across conditions (* indicates two-sided t-test fdr-adjusted p < 0.05 between regenerating/embolized and healthy; boxplot shows the median (center line), 25th, and 75th percentile (lower and upper boundary), and the whiskers indicate the 1.5 × inter-quartile range, with outliers shown as individual data points). Source data are provided as a Source Data file for (h1–3 and j).